Urban traffic suffers from stop-and-go motion that wastes fuel, increases idle times, and causes air pollution. Expanding road capacity can reduce congestion, but building wider roads is not an ideal solution. A more efficient approach is platooning, where vehicles move in close coordination with minimal gaps to increase effective road capacity. Platooning requires reliable communication, and in urban areas with signals, pedestrians, and other disturbances, centralized decision-making becomes important. This paper proposes a cloud-enabled centralized  platooning architecture that leverages the advantages of cloud computing and evaluates how networking conditions influence control performance. The framework is tested in CARLA, an open-source simulator for autonomous vehicles. Two longitudinal controllers were compared, the Intelligent Driver Model (IDM) and a nonlinear Model Predictive Controller (MPC) based on eco-driving formulations. Experiments considered three connectivity setups: local Wi-Fi, private 5G, and AWS(Amazon Web Services) cloud. Metrics included end-to-end latency, fuel consumption, idle time, and average speed. Results show that MPC improved efficiency over IDM, reducing fuel consumption by 18.7% and idle time by 96.5%. The study provides an experimental comparison of MPC and IDM in urban platooning and guidance on the effectiveness and limits of cloud-based optimization.